One AI toolkit forQA engineers
Turn failing screens into evidence. Rough notes into reproducible bug reports. A test session your AI tools can read. One menu-bar app, one shortcut.
The error is on screen,
and it's not selectable
A modal, a stack trace in a terminal, a half-rendered page. Drag a box over it and Khint gives you the exact text, error codes included, then lets you keep asking about the screenshot instead of retyping what broke. The rest of the pass runs the same way: the story becomes test cases at nine, the failure becomes a filed issue at eleven, and the board moves when the fix lands.
Confluence · Team Velocity Gadget
The real Capture & ask window. The region you drag never leaves your Mac until you trigger the run.
Two ways to get from a failing screen to a filed bug
Both end with an issue somebody has to read. They differ on how the evidence gets out of the screen, and on whether the report looks like everyone else's.
Evidence you have to retype
The bug is found in seconds. What takes the pass apart is getting the evidence out of a screen that won't give it up.
- The error dialog isn't selectable, so the message and the codes get retyped by hand, with a typo landing somewhere in the stack trace.
- Coverage is improvised from the acceptance criteria, so the negatives and edge cases depend on how much of the story you remembered.
- Each tester writes a report in their own shape, and the developer reading three of them learns three different formats.
- Filing means a browser tab, then another one to move the ticket when the fix lands, both in the middle of a pass.
Khint, on the region you drag
Drag a box over the failure and the exact message comes back as text. The report's shape lives in a saved agent rather than in each tester's head.
- Extract text pulls the exact message and error codes to your clipboard; Capture and ask questions the screenshot across several turns.
- Select the story and its criteria and a test-case agent returns the happy path, the negatives and the edge cases as a list you can walk.
- The report's shape lives in a saved agent rather than in each tester's head, so every report the developers read looks the same.
- With Jira or Linear connected the palette files it and copies the key. One issue per run, every write behind an explicit confirm.
JiraQA-233The four pieces
a regression pass runs on
A pass is evidence, a shape to put it in, a memory of what build you were on, and a tracker to land it in. These are the four pages behind that, each a shipped feature rather than a roadmap slide.
Capture
A modal, a stack trace in a terminal, a half-rendered page. Drag a box and the exact message and codes are on your clipboard.
Agents
Test cases from a story, then the bug report in your team's shape. The format lives in the agent, not in each tester's head.
Memory
One session per pass. The bug you file at four already knows the build you were testing at ten, and Copy recap writes your standup.
Integrations
Create the issue in Jira or Linear from the drafted report, then Change status walks you through the allowed transitions.
What changes across a regression pass
Three things, and none of them is a browser tab opened mid-test.
- StopsRetyping an error message by handThe stack trace in a terminal and the codes in a modal are text again in one drag, so the repro steps quote what the app said rather than what you recall.
- LandsThe issue, in Jira or Linear, already shapedRepro steps, expected versus actual, a severity suggestion, in one format whoever filed it. One issue per run, every write behind an explicit confirm.
- Gets backThe end of the pass, already writtenCopy recap turns the session, summary and entries, into paste-ready standup notes, and Claude Desktop reads the same live session over MCP.
Common questions from QA engineers
What can a QA engineer actually do with Khint?
Select any text (rough test notes, a user story, a log excerpt), hit Cmd+Shift+K, and run an AI agent on it in place: turn notes into a reproducible bug report, generate test cases from a story, or tighten repro steps before you file them. The same palette captures any region of your screen to text, so a stack trace in a terminal or an error dialog becomes copyable evidence without retyping. Connect Jira or Linear and the palette files the issue too: you never open a browser tab mid-test.
Can Khint write bug reports from my test notes?
Khint drafts from what you observed, not from nothing. Select your raw notes and run a bug-report agent: it returns repro steps, expected versus actual behavior, and a severity suggestion in a consistent format. You review it, then create the Jira or Linear issue from the palette. Because the format comes from a saved agent, every report your team reads looks the same, whoever filed it.
What does a Memory session add to a test pass?
Start a session named after the pass, for example 'Release 2.4 regression'. Every capture, every generated report, and every issue you file lands in it, compacted into a short running context that later agents can see. The bug you file at 4pm already knows which build you were testing at 10am. At the end of the pass, Copy recap turns the whole session into paste-ready notes for your standup, and Claude Desktop or Codex can read the same live session over the bundled MCP server.
Does my test evidence leave my machine?
Only what you explicitly act on. An agent sees the text you selected plus, optionally, the active Memory session's compact summary. A capture sends only the screen region you selected, and only when you trigger it. Memory sessions live in a local SQLite database and are never synced to a server, and your Jira or Linear credentials stay in your OS keychain. Agents can also opt in to PII redaction, which replaces emails and phone numbers with placeholders before the text reaches the AI.
Is there a free plan?
Yes. The free tier includes 300 credits a month, about 10 AI actions a day, no credit card, and all the work integrations are available on it. Paid plans start at €7/month; Pro at €29/month gives 4,000 credits a month. Khint runs on macOS 13+ and Windows 10 22H2+.
Run your next test pass with it
Free with 300 credits a month, about 10 AI actions a day. No credit card. Connect Jira or Linear when you're ready: it takes your own API token.